A multimedia environment for investigation of experimentaland simulated Dough Kneading data sets

I. DELIYANNIS

University of Wales Swansea, Computer Science Department, Singleton Park, SA2 8PP, Wales, UK, csdeli@swan.ac.uk

M. F. WEBSTER

Institute of Non-Newtonian Fluid Mechanics, University of Wales Swansea, Computer Science Department, SingletonPark, SA2 8PP, Wales, UK, m.f.webster@swan.ac.uk

Abstract: A detailed industrially-based multimedia case-study is considered, featuring experimental and simulation dough-kneading data-sets. Multimedia environments are employed, creating flexible implementations, whilst dealing with large anddensely interrelated data. This complicated rheological investigation, aided via multimedia technology, spans across variousparameter spaces. These include geometric, fluid-type, speed, depth and stirrer-shape adjustments. Direct data-comparisonacross valid combinations is actioned, through multi-menus. Voiceover-streams and multiple navigation-paths are characteris-tic features. Multimedia environment-level retiming supports synchronised presentation of non-uniform animations, reducingre-rendering times. Through multimedia environments, direct system-updates are practical, whilst data-duplication is avoided.Such implementation enhances the presentation impact-factor of the work. As such, the system supports intelligent interroga-tion, that itself, may lead to heightened awareness and meaningful interpretation of the data. Whilst the resulting systems fulfilbase-level presentation requirements over various media, the same systems may be used for further data-evaluation.

Keywords: multimedia, human-computer interaction, navigation, experimentation, simulation, stream-synchronisation,content-evaluation.

1. CASE STUDY: DOUGH KNEADINGThe main objective of this work and the Multimedia System (MMS) it has generated are, to illustrate the key factors thataffect the process of dough kneading. This is achieved by creating a multi-media visualisation system for the complete,direct and meaningful presentation of the associated data. Use of MMS can prevent presentation deficiencies introducedwith proprietary software. In addition, the ability to present results comparatively, from case to case, or via interactivemode-switching, is highly desirable. This implementation is a valuable tool for the comprehension of parameter varia-tion, and consequent affect upon the associated flow or process under consideration.

Ultimately, MMS data integration ([11], [23]) has enabled effortless navigation through the vast volume of information,dispensation of data-duplication, and interactive comparison of experimental and simulation animations, within a singleMMS frame. Interaction with animations is implemented through user-interface options. In addition, the MMS may beprogrammed to simulate common interaction modes, as cited in the literature [5]. The flexibility, demonstrated in mul-timedia environments (MME) and rigorous system-design for large data-sets, is based on the uniqueness of the data andtheir interrelationships.

For industrialists, process optimisation may imply significant economy in production costs. Typical products of thedough-kneading process are everyday goods, such as bread, biscuits and pizza bases. Figure 1 illustrates an empty mixer,two different states of kneading (mixer-lid removed), and a typical, final baked product. This is a rich and varied foodsstudy, that of dough kneading, introduced within the frame of Figure 1, [cc.1]. The associated complex free-surfacemovement involves wetting and peeling on vessel and stirrers, which have demanded new modelling algorithms. TheMultimedia System (MMS) trailer quickly scans what was involved [“Show Animation” option, Figure 1]. The mainconcern here, is to identify the factors that affect the industrial kneading and mixing of dough. Industrial mixing deviceshave a number of features that increase their effectiveness, such as dual stirrers and baffles at the outside wall. These in-troduce additional computational complexity into the modelling. The problem has been investigated in a systematic

manner, via dual simulation [18] and experimentation [4] studies, using simplified model-mixers and model-fluids. Theintricacy involved was incremented gradually, as increasingly complex fluids and mixer settings were introduced.

Considerable interest is evident in this applied case-study, from computer-science, rheological ([10], [12]) and industrialpoints of view. From the computer-science perspective, various requirements are exposed. The sheer size of data, intro-duces storage/retrieval limitations, particularly as non-linear access is supported by the end-system. Synchronous displayof multiple streams results in system/network-bottlenecks, particularly as delivery over Internet communication channelsis featured by the end-system. In addition, presentation of the data in a meaningful manner has triggered the need foradvanced and multiplexed navigational routes, which permit various investigative scenarios, alongside pre-determinedpresentation-paths. Research in this interdisciplinary area of interactive multimedia has resulted in the development of awide range of multimedia systems, destined for courseware, research and industrial applications (see [26]).

With respect to visualisation and presentation aspects, novel organisation techniques and state-of-the-art visualisationalgorithms are utilised. Examples include the application of animated visualisation techniques, to display the position offree-surfaces (“Pov-Ray” with spherical-particles, see Figure 21 below), in form. This is accompanied with a range oftechniques to visualise the free-surface in a realistic animated setting, using 2D and 3D views, and solid-modelling pro-cedures. New interface styles for MMS data-presentation have been developed, using a hierarchical organisation, based onmodel adjustment (via customised multi-menus) and navigational graphs, introducing intelligent navigation ([24], [26]).

Figure 1, cc1

1.1 Many-formatted data considerationsAdded complication is introduced as data are modelled, calculated and delivered, by multiple research groups, working inparallel. Each group has employed individual methods, to store and provide the data (variety of standards), providing thedevelopment workload. Experimental and simulation results generated, range over a variety of investigation areas [Fig-ures 7 to 11], featuring a large number of variables. Frequently, calculations may be repeated with algorithm adjust-ments. Thus, each time data are received, increased complexity is introduced. It is necessary to identify unique, fresh datato be categorised and inserted, appropriately. Categorisation, ordering and duplication problems arise with PPT and newdata. The above deficiencies are amplified, as lack of data-format standards introduce further conversion complexities.Within MMS, new data are inserted, or replaced, straightforwardly to the system without-preferred order. Some initialpractical problems to be tackled include the collection, categorisation, visualisation and presentation of simulation re-sults, having established data-sharing practice. This minimises data conversion, and as a consequence, the visualisationworkload. The plethora and diversity of data, demands MMS techniques, relating to efficient content-updates and directdata-insertion.

Suitable interaction mechanisms had to be created, able to cope with frequent data updates and content re-organisation.The final result involves interactive combination of large volumes of experimental and simulation data, under a point-and-click integrated environment. Experimental data are presented comparatively, either with simulation data (directly)[Figure 15], or via appropriate indirect links [Figures 16, 17]. Comparative results in animated stream format are used.This data format facilitates multiple modes of interaction, via parameter-adjustment. Such combined and interactive fea-tures, render the use of a synchronised and ordered, interactive presentation mode, highly valuable, particularly whencomplex, multi-type data are to be assessed.

The problem was modelled in two and three dimensions [8]. For fully-filled simulated cases, results were generated onfive separate horizontal slices through the vessel at equal distances set apart, from bottom to top. The free-surface posi-

tion was calculated separately in part-filled cases. Customised “Multi-Menus” were required for these large data sizes, anddue to the limited navigational ability of available commercial software. A necessity is to navigate in a user-defined or-der, whilst simultaneously, allowing direct-access to secondary information. Evaluating the data, or attempting to inves-tigate flow behaviour through variable-adjustment, is a demanding task, for linear presentation organisations. Standardpresentation tools are designed to present data by default in a pre-determined manner, disabling direct interaction betweenmultiple variables. Here, utilisation of interactive techniques is required, to allow smooth navigation through data-sets,enabling parameter adjustment, on demand. This MMS allows for such direct adjustment of variables, and the presenta-tion of interrelated data, under an integrated environment.

Streaming problems were also encountered, as once completed, the size of the compressed presentation was large, suffi-cient to fill a Compact-Disk storage-medium (CD, holding over 600 Megabytes). Stream-optimisation and re-use featureswere required, to reduce data-bottlenecks. For example, all multi-menu icons are pre-loaded into memory before theMMS commences. They are not unloaded until the MMS is closed, reducing loading/unloading times. The cruise-control path (described later) is optimised, preloading (caching) the contents of the next slide, whilst at previous slides.

1.2 Content-based user-interface designThe Multi-Menu structure acts as a graph that depicts link-availability. So, for example, when a geometrical setting isselected, the Multi-Menu provides several possible navigational routes. Interaction is elevated considerably, via the in-corporation of such graph-structures within the interface. It becomes immediately apparent to the user, that the nodes ofthe graph are related to the associated settings. Upon selection, the Multi-Menu constructs (referred to as “cast” mem-bers), are programmed to adjust their visual properties (Visible, Dimmed, Selected, Non-Visible), using modern graphi-cal user-interface conventions ([14], [15]]). This property provides direct user-feedback indicating the current choice, and acomprehensive list of other link-paths.

Navigational and user-delivery problems became apparent when the MMS was used in practical presentations to differenttypes of audience. Data complexity leads to complicated navigation paths in presentation-mode. This problem was over-come, using multiple colour-coded “cruise control” mechanisms, predefined orientated-routes through the main MMSsections. In this mode, the user is permitted to interact with the data in a non-predetermined manner, and to toggle backto a predetermined mode, as desired.

PPT to MMSTypical organisation problems cited [7] were overcome by first evaluating the data under consideration, and consequentlyprogramming interaction, using the MME to support systematic categorisation and interconnection of the data. The re-sulting system does not rely on external links to access detached presentation units: instead it integrates the data in itsoriginal form. Data duplication is removed, by using various object and MM techniques. The data are encapsulated andreferenced within the so-called “cast” file. This enables multiple (internal) data-instances to be referenced to a unique (ex-ternal) file-object. For each MM-frame requested, data are dynamically called from the “cast”. Operations such as resiz-ing, rotations, transformations, animations, fading, colour or frame-rate changes are applied, before the MM-frame isdisplayed. For subsequent frames, data-instances, already processed, can be retained on screen. New instances follow theabove procedure. The data are stored once within the MME, and accessed from various related MM-frames, on-demand(multi-sorted data-context). This integration process does not require the utilisation of external data-format conversionalgorithms. Figure 12 below, illustrates how the MMS incorporates requested data within the “Main Presentation” frame.Some of the advantages that this implementation demonstrates are immediately apparent. Duplication is removed, anddata integration/categorisation guarantees, that all related frames can be accessed in presentation mode. Insertion of newdata into the MMS, performed via the “cast” within Director, does not require a specific order. Nevertheless, grouping oflike-data units rationalises frame-linking, that establishes the interrelationships between data.

MME may be programmed, so that automatic data-validation is performed for the necessary running components. In thismanner, missing content-components may be recovered from alternative locations, prior to launch. Otherwise, diagnosticmessages may be displayed, to describe the nature/type of discrepancy, the appropriate action to be taken, and the statusof the system to proceed. This procedure guarantees system-integrity and robust status, re presentation. Traditional pres-entation software, such as PPT, may only achieve presentation integrity via manual verification.

Adjusting Mixing-OrientationTwo mixer orientations are modelled, simulated, and validated by experiments within the framework of this project.First, dough-mixing was modelled in vertical-orientation, and subsequently horizontally, for fully-filled settings. Be-yond fully-filled scenarios, free-surface situations were considered, under both settings. In this manner, a deeper apprecia-tion may be grasped of the fundamental aspects at play within these processes. Choices of mixer are provided, typicallyin Figures 3, 12 (mixer-geometry instance), and Figure 2 (vessel-orientation choice), taken from the Multi-Menu.

Figure 2 Figure 3

Adjusting Mixing-ComplexityMixer geometry itself, consists of four designs of increasingly complexity: single-stirrer concentric, single-stirrer eccen-tric, double-stirrer eccentric, and double-stirrer eccentric, with baffles on vessel-wall. These settings are accessible usingthe associated menu (left-screen), [Figures 12, 13, below], demonstrating transition between two such instances. Vessel-orientation and free-surface results can be accessed, via the previously described bottom-screen horizontal set-of-fouricons. Navigation is actioned through icon selection. When passing from one menu-option to another, a transition be-tween graph-nodes is performed. The MMS is programmed to respond to these transitions by adjusting the Multi-Menu,highlighting links for further user-interaction. As a consequence, the system responds automatically to subsequent menu-choices, for which inter-related data are held (for example, a geometrical adjustment), and displayed directly, by default.Repeated use of this feature across the MMS, eases navigational strain for the user, aiding direct comparison betweendata-instances. That is, without utilising complicated control commands to reach disparate data. In instances where, nei-ther simulated nor experimental data are available, the user is directed to the principal menu for that specific choice.MMS specification is orchestrated through a graph data-structure, supporting multi-media data, frames and links. At bot-tom, right-of-screen, icons are positioned and colour-coded to indicate specific displayed-variables.

Using User Interface techniques to improve InteractionUser-interface conventions were employed during MMS development ([21], [16]). Colour and link-intensity settings fol-low modern standards and are consistent across frames. Layout choice accommodates the complex structure of data andmenus, allowing for maximum viewable area. In some instances, link-icons are blended alongside the data. This enablesidentification and easy user-access to data-related links. In addition, this can help deal with instances where more thanone choice is provided for the same parameter. Red colour is utilised to indicate selection (free-surface and speed-variation), as icons used are too small in size, and higher visibility is required to depict selection. Here, a non-strict ad-herence to user-interface guidelines is utilised [15]. These conventions are adopted, whenever toggle navigation-menus areprovided, as standard in modern user-interfaces. This factor enhances user-familiarity and experience with the MMS, itsmulti-faceted data-content and powerful navigational capabilities. Additional interface-enhancing graphic properties havebeen employed. These allow the MMS to adjust link-options at each stage, and comply with modern user-interface stan-dards, utilising colours, informative graphics and audio feedback. The importance of consistency across the user-interfaceextends to interaction. A well-designed interface can be rendered useless, without valid and predictable interaction. It istherefore imperative to design the interactive mechanisms and MMS content-linking appropriately. This is achievedthrough the sound use of graphs, where frame-linking can be tested for completeness at any developmental stage.

1.3 Multimedia – “Multi-Menus” and organisationThe MMS enables user-interaction at multiple levels. As indicated earlier, the order of interaction, through multi-menus,is first the geometrical level (geometric adjustment), and then, at each subsequent level across the various relevant vari-able adjustments. The ability to create and re-program each menu at every instance, is one of the most powerful featuresoffered by the MME. Multi-menus are designed to accommodate interaction in multiple modes. Multi-menus are a com-plex construct, both to program and for the MMS to support. So, when should this interaction mechanism be used?Multi-menus were avoided, for slide-sorters, relating to presentation organisation, instead of actual content. An educateddecision must be made, to link only data that can be categorised into sub-sections that does not apply to the wholeMMS. Further justification for multi-menu use, is that interactive modes of the presentation are best represented throughgraph-structures. [6], presents a detailed account of the different links that one may incorporate, and the limitationstherein of standard software today.

A multiple menu instance is illustrated in Figure 3, together with its corresponding graph-structure. Such menus cater fordynamical adjustment of data-content. Currently-selected nodes are indicated using larger circles. “C on bars” representscomparison between two, or more, related data-sets (forming a node by itself). The user can compare data-instances, in-teract with the information presented (via mode-switching), and hence, smoothly navigate through the data-tree.

MMS organisation follows the project plan structure. This facilitates data integration, as the same structure was utilisedfor the creation of the .ppt. It also enables the identification of duplicate data, and allows for rapid content updates. First,additional data are appended to the existing .ppt files, and subsequently, to the MMS. Multiple modes of organisationcan be observed within the MMS, each at a different level. Starting from the programming level, MM-frames are organ-

ised into related sets. The sets themselves are not ordered within the MMS. This would create insertion, or movement,of MM-frame sets. Colours have also been employed, to indicate alternate frame-sets. Interaction follows the graph or-ganisation which, in turn, is based on the project-plan structure. Additional interactive features include: slide-sets pre-sented using a combined slide-sorter mode; alongside actual slides; or use of “next slide” button. These are accessible,both from the main-menu, or from related frames. Such a modular organisation at each MMS level, allows for the im-plementation of the multi-menu structure, as individual components with unique properties amongst the separate sec-tions. Thus, the same graphical icon can have multiple functional-response, across different frames, enabling optimalprogrammability and adjustment, according to the data.

The interactive mode accesses the full power of the MMS. A panic-button is included top-left (Dough Mixing INNFMRTN) that returns the user to the starting point. “Variable-priority” rule applies: once a variable has been selected upon amulti-menu, switching to a second multi-menu, will automatically inherit prior variable-choice. By default, this enablesdirect data-comparison between frames. The user can access ‘green’, ‘orange’ and ‘red’ cruise-control routes, at will. Asthe primary mode of interaction is full-mode, by default, the user is permitted maximal choice, with combined use ofordered, interactive controls. The availability of on-demand links and the plethora of available routes are strong featuresof the MMS.

A second user-option is to follow a pre-determined order of presentation. This is a desirable feature for public presenta-tion, usually in a limited time-frame and under duress. Under such conditions, large volumes of data can create naviga-tional difficulties. The “Cruise Control” mechanism enables implementation of pre-determined paths, complete withvoiceover (VO). Multiple paths can be created, indicated using appropriate colour-coding. The DK-MMS utilises threeseparate control mechanisms. ‘Red’ guides the user to the main MMS sections, ensuring all major sections are accessed.‘Orange’ button is a pre-defined, more exhaustive, navigational path through the MMS. ‘Green’ cruise-control is a par-ticular elective path, with voiceover clips, explaining underlying content.

1.4 Graphs to model MMS structure

Figure 4 Figure 5

Figure 6

Within the current case study, extensive use of graphs and related representation structures have been utilised, for therigorous implementation of navigation, using interaction. The main reasons for this are: large volume of presentation

material; complexities with “Multi-Menus”; need to create a MMS map defining positions for further data-inserts; valida-tion for presentation mode specification.

The type of graph used with this case study is a multi-connected graph1 [Figure 4]. For the organisation of PowerPoint(PPT) presentations, a dynamic linked-list structure is employed [Figure 5]. Each branch is itself a presentation, linkedto external data (such as experimentally-captured and simulation-generated animations). One immediate disadvantage ofthis organisation is that, when for example, speed and depth information is demanded, the Z2-plane of the fully-filledcase must be used twice. This duplicates the volume of information stored within the PPT, thus increasing final presen-tation-size. The above data-categorisation is followed within the multi-media organisation, but in a more liberal manner.Replacing strict-ordering of frames (as in above structure), data are retained locally to one instance in the MME. Linksjoin the top-level to the case-study introduction, flyer-animation and project-plan slide-sorter. The second-level (indicatedby dark-grey intensity), consists of the menus organised across geometry and setting-complexity (one-stirrer, concentricto two-stirrers, eccentric with baffles). Subsequent levels branch out from the second-level options, providing links toindividual frames and their related information.

Figure 5 shows a section snapshot of organisation for links to MMS-components. The columns represent some of theimage components, animations and sounds (used to construct Figure 17, below). Adjacent columns represent relatedMM-frames. To represent this hierarchy using a graph, each node must represent a MM-frame. Each such node can inheritproperties/attributes from another; new attributes/behaviours may be appended, or replace those previously. This organi-sation is adaptive and flexible. It may match the relationships for any given data-set. Attributes need not be programmedfor each frame, being inherited for a copy-object of an existing frame. When an attribute is programmed, it generates anew database object that may be associated with any component of a frame. Frame components are stored uniquelywithin the database and are referenced to the same item, when multiple instances are on display. Instead, a link is em-ployed, if many instances of the same data element are required. The user may select when the memory is flushed of thatparticular element.

2. GREEN CC-TOURTo provide some insight into the data-content and the power of the MMS, details are provided for the green cc pre-selected route across the data-domain. Two principal requirements are fulfilled: first, expert description of the data is de-livered through interactive MMS access. At the same time, it is demonstrated how this is achieved using newly-developed presentation methods. Note, that this navigational path is not restrictive, and can be overridden at any point,via selection of a multi-menu link.

Project overview

Figure 7 Figure 8

The project involved: Horizontal or Vertical Mixer Orientation; Incremental Stirrer Complexity for One Stirrer Concen-tric, One Stirrer Eccentric, Two Stirrers Eccentric, Two Stirrers Eccentric with Baffles; Viscous and Viscoelastic Fluid-types; Speed of Kneading of 12.5, 25, 50 and 100 rotations per minute; 3D Investigation of the flow across depth formultiple planes; Fully-filled and part-filled mixing; Free-surface in two and three dimensions, including transient devel-opment of the free surface; Stirrer shapes that include Full-Stirrer, Half-Stirrer Horizontal, Half-Stirrer Vertical ([19] [20]) 1 VF, HF, VP, HP indicate access to and from Vertically Filled, Horizontally Filled, Vertically Part-filled and Horizontally Part-

filled geometries, respectively. Z0 to Z5 indicate intersecting planes, sliced through the vessel. N, IN, VE indicate Newtonian,Inelastic and Viscoelastic fluid choices. VR, SR represent rotation-type, Vessel Rotating or Stirrer Rotating. R1 to R8 areReynolds Number values, indicating Speeds of Rotation (12.5, 50, 100, and 200 rpm, respectively). C is a combinative-nodeoption, for each of depth, speed, fluid or rotation-type considerations. All links to external sections are suppressed withinsub-graphs. Also, all nodes within a sub-graph are linked forming an internally fully-linked sub-structure.

[“Project Summary” option, Figure 1]. Slides showing the extent of analysis are displayed [Figures 7 to 11]. The re-search plan summarises the simulation and experimental tasks undertaken within the remit of this case study. With simu-lation tasks, the intricacy of investigation is amplified, as setting adjustments draw closer to real-life situations. Initially,results were generated for model fluids, modelled in a two-dimensional setting, and simplified mixer-geometries. Furtherinvestigations included three-dimensional and free-surface modelling. The simulation results at each stage, were validatedagainst those generated from actual kneading experiments. Fully-filled and part-filled cases were considered, in both ver-tical and horizontal mixer orientations. In industry, the horizontal setting is used for biscuit making, and vertical forbread (production) kneading

Experimental fluids cover [Figure 7]: model-fluids (for transparency), model-doughs (for material time-independency), toactual industrial doughs. Each of the above considerations consists of various geometrical and parameter adjustments ofincremental complexity [Figure 8].

Fluid model representation ranges from Newtonian to inelastic, and finally, to viscoelastic (viscous to viscoelastic) [Fig-ure 9]. Adjustment of geometry and modelling complexity transcends from one stirrer, concentric to the final case of twostirrers, eccentric, free surface [Figures 9 and 11]. Incrementation through experimental studies is outlined in [Figure 10].Numerical modelling considered the following scenarios: Viscous Filled (2D, 3D); Viscous Part-Filled (2D, 3D); Vis-coelastic Filled (5 Fluids, 1 Stirrer, 2 Stirrers) [Figure 11]; Viscoelastic Filled (Full Stirrer, Horizontally Shaved Stirrer,Vertically Shaved Stirrer) (not shown).

Figure 9 Figure 10

Figure 11

Five industrial companies provided industrial trial data and the interdisciplinary research cross-referenced INNFM ex-periments at Aberystwyth with modelling/visualisation at Swansea. The main aims were to: analyse the stirring ofdough; gather information on mixer-design choice; relate this against dough rheology; predict how to maximise stretch-ing work input to the dough and enhance the build-up of material structure (i.e. kneading). Kneading, is itself a key as-pect to the overall manufacturing procedure.

The MMS accommodates experimental and computer-generated animations, including synchronised graphs with anima-tions. In addition, the diversity of data and lack of standards had to be resolved, when animated streams are generated.New MM mechanisms, designed for synchronised presentation of animated data, were sought for this case study. A fur-ther highly-desirable MMS characteristic is the ability to navigate through the data on-demand, using minimal steps toreach the desired information. This demands investigation into the data complexities2, to program interaction routes intothe MMS that fulfil minimum path-length criteria. The use of graphs, met the required topological representation, ena-bling optimal navigation and interconnectivity across data-content. A variety of abstraction levels were considered: frominteraction between major data-sections, to interaction at the deepest level among data-instances.

2.1 Accessing the dataPerspective static views of flow patterns are illustrated for a filled, one-stirrer mixer, with anti-clockwise vessel-rotation,shown half-way up the mixer at 50 rpm (a standard speed and model inelastic fluid) [Figure 12, cc.2]. Asymmetricalstructure is apparent with an off-centre vortex: pressure, shear-rate and rate-of-work extrema are localised to the stirrer. Asabove, each menu icon is, in fact, part of a programmed network (a graph) of the presentation (covering variation inspeed, height, material and rotation-type). This case can be contrasted against the two-stirrer instance, upon user-selectionof the two-stirrer geometry [Figure 13]. The mixer-geometry is sampled with two-dimensional horizontal slices indicatedas Z0 [the base of the mixer] to Z5, the lid of the mixer, with equal separation gap between each, and all normal to thestirrers. Variation across depth is studied through a comparison between results across planes, Z0 to Z5. The defaultplane of reference is the central Z2-choice; idealised to 2D and removed from end-effects. Interrogation of depth-variationreveals the regional changes encountered within the mixer. This casts light upon localised field variable distribution.This renders a deeper understanding of the flow and process as a whole, and demonstrates the three-dimensional aspects.Some symmetry is observed about the stirrers and a central figure-of-eight vortex pattern emerges.

Figure 12, cc2 Figure 13, cc3

Depth-wise investigation

Figure 14, cc4

2 Data complexities are defined in this context as the set of relations between MM-frames. Each frame containing a set of cast

members (multimedia or static data-types), is classified as a single data-instance.

Three-dimensionally information, enable appreciation of depthwise-distribution in rates of shear, extension and work,against stirrers and lid [Figure 14, cc.4]. Here, 3D data for the two-stirrer configuration are displayed. Pressure variesmainly about the stirrers; deformation in shear and extension, likewise, are localised to the stirrers; shear-rate maxima atthe lid, extension-rate at the lid and vessel-bottom; rate-of-work done follows the shear deformation, being localisedwithin the narrow-gap section between stirrer and vessel-wall, towards the lid. In the two-stirrer case, the result is sym-metrical and localised close to the stirrer. A characteristic animation state can be accessed by selecting the arrow next tothe depth-options. Minimal change is reflected in streamline-patters, indicated using animation for one-stirrer and two-stirrers, through depth.

Synchronisation: Simulation and ExperimentAn advanced MM feature offered, is the property of data synchronisation, that may be achieved in a variety of ways. So,for example, animations may be synchronised to present information, comparatively. Extensive use of this feature is evi-dent throughout the MMS. Every effort has been made to present information in a concurrent mode, wherever possible.Generally, animation re-timings [3] must be introduced, to synchronise concurrent clips. This method is not restrictive,as other implementations are cited in the literature [17]. The objective is to present multiple animations, so that at anygiven instance, the information provided relates to the same parameter. The parameter can be any attribute within theparticular case study. For example, from the layer used (observing the flow through depth), to speeds (where all clipsmust be provided for the same speeds). To achieve this objective, control over streams is required, in terms of frame-rate,precise timing to stop/start and control over displayed size. Comprehension of display is aided by synchronous labelling.

2.2 Speed-wise interrogationAnimated views, passing through increased speeds, allows the direct cross-reference of simulated data, in pressure, exten-sion-rate and motion-blur fields, against experimental flow visualisation (bottom-left, stirrers indicated) [Figure 15,cc.5]. The correspondence in vortex structure is striking. Motion-blur clips at four, set speeds of vessel rotation [Figure16, cc.6], identify the twisting of the vortex structure with increased speed (in the direction of rotation). This is corrobo-rated in high-speed camera, laser-scatter stills of 1% cmc fluid [Figure 17, cc.7].

Figure 15, cc5

At 50 rpm, the motion-blur flow patterns between one- and two-stirrer mixers may be contrasted (on the left), whilstalso taking the industrialist’s view with stirrers rotating (on the right) [Figure 18, cc.8]. The variation of mixer speed isgiven under Reynolds number representations of 2,4,8 and 16, which translate to speeds of 12.5, 25, 50 and 100 revolu-tions per minute, respectively. Mixing-speed is highlighted, within the MMS, at the bottom of the “Multi-Menu”. Byselecting the individual speed-icons, streamline simulation, experimental laser animation, pressure animation and staticthree-dimensional extension-rate results are presented, simultaneously [Figure 15, cc.5]. The streamline and pressuresimulation results, are animated through speeds, thus the information presented, steps synchronously through the varioussettings automatically. Typically in PPT, external applications must be used to display individual animations and syn-chronisation is barely possible. Precise frame-based PPT animation synchronisation can be achieved, through renderingof the individual streams into a single stream. However, this is an exhaustive procedure, both in terms of time and com-putation.

Simulation versus Experimental DataA variety of techniques were used, for the collection of experimental data, ranging from measurements to identify fluidsurface position, to full flow video-footage. The allocation of these experimental data has been in comparative mode tosimulation results (taking precedence), for immediate cross-reference. In some isolated instances, experimental resultsalone are presented. The interface is designed to allow the immediate comparison of simulation results through the“Multi-Menu” construct. Additional interactive buttons are provided, wherever this mechanism is less than adequate. Anexample is the experimental speed-adjustment results [Figure 17], demonstrating the “figure-of-eight” streamline pattern

visualised, and its change of orientation as the speed increases, for the two-stirrer eccentric case. Figure 16 [cc.6], depictsthe comparatively organised motion-blur results when speed increases from Re=2 to Re=16 (25 rpm to 200 rpm). Themost interesting flow characteristic that can be observed with this image organisation is the rotation of the figure-of-eighttowards the stirrers, as the speed is increased. Upon comparison of the simulation figures with the experimental images(versus Figure 17, cc.7), the same qualitative trends are observed. Note that, the orientation of the stirrers is not alignedat the same position in both experiment and simulation.

This far, the vessel has been rotating and the stirrer is fixed. This approach is more straightforward to model, demandingless computational effort as re-meshing is avoided. Industrial-settings usually involve the opposing scenario, a moving-stirrer and fixed-vessel. Therefore, numerical inversion of the process was performed to verify that under these rotation-conditions, simulation results are close to experimental-type flow-characteristics. Figure 18, [cc.8] displays an animatedframe, presenting both settings comparatively. The results follow physical expectation.

Figure 16, cc6 Figure 17, cc7

This comparative multi-media feature is powerful. It enables direct parametric contrast and data-evaluation. Equivalentfunctionality could only be accomplished with propriety software (PPT), by sequencing the information in-order (copy-ing out of context), or via slide duplication. Either option, would result in considerable increase in PPT file-size.

Figure 18, cc8

2.3 Free-surfaces: mixer-speed variationWithin other instances of the case study, for example “free-surface, one-stirrer”, various speeds of rotation can be selected.The selection refreshes the surface animation sequences and the graphical animation, with the appropriate data. The con-sequence of this presentation method is that the programmed system, displays the information in on order dictated by theuser. This is convenient when comparisons are made between settings, or across instances.

Next, we turn to the vertical, part-filled instance (for bread-making), with a central-stirrer, vessel rotating and compare thefinal-rise, surface-position graphically against experiment. Agreement is encouraging. Similarly, we may combine caseswith three set-speeds, 25, 50 and 100 rpm, to demonstrate variation of fluid height-rise at the outer-vessel. Such resultsare obtained by modelling the peeling-off and wetting-onto the surfaces, via the adjustment of fluid-surface line-segments, according to their stretch and angle from the solid boundary. Relief of limiting stretch, also relieves criticalboundary stress-levels.

Figure 19, cc9 Figure 20

Figure 21, cc10

These figures demonstrate the power of interaction across settings. Figure 19 displays the 25rpm setting. Figure 20 pro-vides the 100rpm speed-setting data, with transient free-surface profile evolution as recorded in line-graph form. Clearly,one can observe that the higher speed-setting, generates greater rise in the fluid free-surface. The animated nature of thedata enables the viewer to appreciate that, the vessel is rotating and not the stirrer. It is not necessary to indicate direc-tion/driving motion, as required in static format (journal papers/reports).

Within figure 21 [cc.10], comparative visualisation of the simulated, steady-state free-surface profiles, contrasts the re-sults, for various speed-settings, with the experimental measurements charted in graphs below. When considering part-filled cases, the free-surface was modelled in both vessel-orientations, horizontal and vertical. Modelling and computervisualisation was an integral part of the research conducted, demanding new combinations of techniques to visualiseflow-surfaces.

The experimental and simulation studies generated concurrent results in many instances, for flows of identical setting andvariables. This enabled comparison. When experimental and simulation results are presented simultaneously, it is im-perative to ensure that the information is compatible. An example where comparison of surface-line profiles illustrates theproximity of the simulation results to their equivalent experimental measurements is described below. Close observationreveals that the levels of fluid were not identical, and the shape of the lines reflect a higher Reynolds number for the ex-

perimental case [Figure 21, Simulation Re=16, Experimentation Re=18]. This is best demonstrated at the maximumspeed of 200rpm, in the bottom, blue lines.

Figure 22 Figure 23, cc11

Figure 24, cc12 Figure 25

The resulting particle-seeded images from the visualised simulation data are displayed in Figure 22 (three-dimensional,steady-state, axisymmetric, rendered animation model). This is viewed from a fixed angle, towards the centre of the ves-sel, demonstrating the identical surface characteristics as those in the experimental image.

Algorithmic details to impose wetting and peeling involving free-surfaces, is covered in figure 23 [cc.11]. Here condi-tions can be allocated to sets of fluid-particles that make contact with the surfaces, to represent and control both peelingand wetting. A more complicated vertical scenario is that with a single, eccentric-stirrer (vessel-rotating). A surface trian-gulation (heavy on graphics) illustrates the complex shapes encountered [Figure 24, cc.12]. Different viewing angles,with lighting and shading, indicate the surface rise ahead of and dip behind the stirrer. These validate our predictions.

The four simulation images [Figure 25], taken at different viewing angles, can be contrasted against the experimentalsingle-speed of 150 rpm. The experimental views of Figure 26 (still photographs), cover a single viewing angle at fourspeed-settings, for corresponding steady-states. There is increased contortion of the surface as speed gathers. Similar flowcharacteristics can be observed in both sets of static images. Some volume of material is located in front of the stirrer andtwo almost symmetric dips are located at the backside of the stirrer, towards the centre of vessel. The experimental build-up of surface structure is more clearly conveyed as animated from a rest-state at 250 rpm vessel-speed. This flow video-footage is instigated by mouse-click upon the central-frame of Figure 26.

In contrast, horizontal mixing (used for biscuit-making), may be viewed from one end at four different times [Figure 27,cc.13]. Here, we detect wetting/peeling at the outer-vessel and peeling from the stirrer, as time progresses. First, we viewthe simulation through an animation clip [Figure 28]. The wetting/peeling at the outer-vessel is a dominant feature. Thismay be contrasted against the corresponding experiment for a syrup at 50 rpm [Figure 29]. The surface attachment struc-ture around the inner-section of the stirrer and the central, flat surface-shape are finer detail to capture. Even these particu-

lar details may be predicted, by careful, localised adjustment of control parameters (a) for the inner and outer stirrer-sections (left-image, constant factor; right-image, dynamic setting, Figure 30).

Figure 26

Figure 27, cc13 Figure 28, cc14

The corresponding results from the simulation and experiment, reflect close agreement. These animated clips allow theuser to compare the similarity in evolutionary development and final position of the free-surface (i.e. in time). With theone-stirrer eccentric case, the complexity of visualisation increases, because there are no symmetries within the associatedimages. Video-footage of the experimental data (available for three speed-settings: 25, 50 and 100 rpm) is activated viabottom-screen button-selection and displayed in the central window, Figure 29, conveying fluid surface movement intime, as the vessel is rotated from a flat rest-state. The fluid is Newtonian in these instances. Figure 30 [cc.16] displaysthe free-surface position when peeling conditions are adjusted about the stirrer. Adjustment in a-value, and the corre-sponding visualised result for previous and final setting, are presented comparatively.

2.4 Fluid-type comparisonThe range of materials is broad, varying from inelastic to viscoelastic test-fluids to real-life dough. An example of acombined-menu instance (left-screen) is presented in Figure 31. The menu-choice illustrates (central-window) the corre-sponding data relating to fluid-type variation. The fluids under consideration are Newtonian versus Inelastic, observed onthe Z2-plane for the fully-filled variant, with vessel-rotating at 50rpm. This is equivalent to the setting of Reynoldsnumber 8. The frame-title reflects the data-type being displayed.

Figure 31 Figure 32

Viscous SectionBy selecting the “Newtonian” button, the MMS directs the user to the “torque versus time” graphical comparison. Thisillustrates contrast between Newtonian and inelastic materials, for the 50rpm setting. It is the only such data-set availablefor Newtonian materials; hence, selected by default. When the one-stirrer, eccentric button is selected next, correspond-ing, “torque versus time” one-stirrer data is provided (variable-priority rule applied).

An additional link is provided with the cross-referencing arrow, under the “Newtonian”, “inelastic” and “viscoelastic”buttons. The arrow menu-choice is employed to cross-reference information relating to two, or more cases. Appropriatedata are presented within the central window. This enables the comparative presentation of results, for example acrossgeometries or materials. Figure 32 contains a comparative data-combination example, with one and two-stirrers, eccen-tric, torque versus time trace and Motion-Blur animation.

2.5 Viscoelastic Analysis2D viscoelastic results are available for one and two-stirrer cases, and for three-stirrer geometries: full-stirrer, half-horizontal and half-vertical. In addition, data have been computed using parallel processing ([1] [2]). General presentationslides, related to the parallel algorithmic implementation, can be accessed from within the viscoelastic section, and themain menu. The viscoelastic section can be accessed from both one and two-stirrer, multi-menus (viscous section). Vis-coelastic results are also compared across different model fluids and geometries. The comparison is enabled, upon selec-tion of the “Viscoelastic” option within the “Multi-Menu” construct; or alternatively from the “Viscoelastic” icon [Figure31, one or two-stirrer eccentric, viscoelastic option]. Results for filled, one-stirrer and two-stirrers, eccentric cases aredisplayed directly, for variables of shear-rate, pressure and stresses (radial stress t rr, hoop stress tqq, and shear-stress t rq).The results range across five viscoelastic fluid-models: LPPT (0.02), LPPT (0.025), EPTT (0.02), EPTT (0.025) andOLDROYD-B. The notation implies Phan-Thien/Tanner models (PTT), of linear (L) or exponential (E) form. The MMSallows direct-access to any stirrer setting, model and variable, via the use of the bottom-screen-menu with icon-buttons.This mode of interaction enables the selection of different models, whilst results for selected flow variables are displayed.

One and two-stirrersHaving entered the viscoelastic section [Figure 33, cc.17], navigation across five model fluids is provided. With the useof the local multi-menu options, to switch between model fluids, one may compare results across component stresses,pressure and shear-rate for all five models. A link allows direct-access to all five-models for increase of Weissenbergnumber, using a tree-like structure. At the top-level of this section, without model selection, general information relatingto all five models is displayed. Upon fluid-model selection, rheological properties for that particular fluid are presented.A further choice of variable is required to display data from a palette of icons. With a selected model, material functionplots of ms(g) and me(e) appear. In addition, links may be activated to further field-variable data inclusive of: Shear-rate

(

†

g•

), rate-of-work-done (rwd), pressure (p), radial stress (trr,) hoop stress (tqq), and shear-stress (trq). Within these MMframes, the data are presented in three or four-up format, with increasing Weissenberg number. “Comb-1S “and “Comb-2S” buttons allow direct interaction between one and two-stirrer results, for all PTT models of the study. Data in bothvisualised and tabular format, are available within the slide-sorter. At each child-frame, one may toggle between one- andtwo-stirrer information. The data presented will remain of the same type, switching to selected geometry. The “We Var”button links to data for only the EPTT (0.25) model. Component stresses, pressure and shear-rate data are available, withvariation of Weissenberg setting in the range of values 1 to 3.

Figure 33, cc17 Figure 34, cc18

The two-stirrer section maps the organisation adopted for one-stirrer. [Figure 34, cc.18]. Hence, the utilisation of theMMS allows interaction with the data selectively, and more importantly, across cases. This interface feature can be usedto demonstrate the difference the introduction of a second stirrer in mixing flows can have for the materials under consid-eration. Data in tabular format are also accessible for one- and two-stirrer geometries [Figures 35 and 36, cc.19 andcc.20]. These display simulation measurements of shear-rate, local rate-of-work and power. Tabulations of localisedwork-input for the double-stirrer case, reveal that elastic work (stretching, shown in red) is dominated by viscous work(shown in blue). Here, shear influences prevail in the totalled work-input. In contrast, the asymmetric, single-stirrerdesign provides ten-times the elastic to viscous work: this is amplified for fluids with some strain-hardening (as occurswith dough). At this point, we arrive at a crucial observation: optimal kneading for dough is achieved with more asym-metrical mixer designs, one-stirrer better than two. More complex stirrer shapes are usual. In this respect, we observe forrealistic dough that flat-bottomed, half-stirrer shapes, produce better results (see below).

Figure 35, cc19 Figure 36, cc20

Stirrer-shape adjustmentAt the final point of the cc tour, attention was given to more dough-like, viscoelastic materials, and filled scenarios. Afurther aspect of investigation within the Viscoelastic section, is that concerning stirrer-shape for those more dough-likematerials. In particular, contrasting how this affects the flow, and ultimately, the process design being modelled. Stirreradjustment appears in the “mixing with two-stirrers and baffles” section, hence the selection-icon is appropriately posi-tioned and designed to reflect the available link. Three stirrer-shapes have been considered: full, horizontal-shaved andvertical-shaved [Figure 37, cc.21]. Here, we may observe the stretching and shear stresses across the mixer for a single-stirrer design, or one with a double-stirrer. Maxima in stress are localised about the stirrer, in the narrow-gap betweenstirrer and vessel; the hoop stress dominates.

Three menu sections can be utilised, each with different organisation of the data. The first menu is selected already,within the stirrer-shape, adjustment section. It consists of three slides that present general model information, and thebottom-right corner-palette of buttons can be used to present a range of further choices. Again, individual data-instancescan be accessed using the icon-options. This time, results are displayed in comparative view across the three stirrer-geometries. For a selected variable, data for Newtonian, inelastic and viscoelastic fluids are accessible upon selection of a

different setting from the multi-menu. The buttons “E-F1S”, “E-H1S” and “E-V1S” allow direct-access to the results foreach stirrer-shape individually. The results are organised into sets of visualised images. One, that combines shear-rate,pressure and rate of work-done for the same stirrer-shape. Another, for component stresses, and additional MM-frames,that group together numerical results in tabular form, streamline and colour-filled-streamline-plotted images. These are allaccessible upon selection of the bottom-right link-palette, programmed to reflect the data within the current presentationmode. The third menu enables the comparison of streamline results for a given stirrer-setting, across different fluids. It islocated next to the button-palette, and consists of three slides. Each with three Weissenberg number settings, across New-tonian, inelastic and viscoelastic model-fluids.

Figure 37, cc21

3. COMBINING DATA TYPESThe visualised data units, require commonly available organisation methods, for easy identification and referencing. Ex-ternal (PPT) software was used, enabling explanatory data to be inserted directly. Sending the generated file (with all theinformation embedded), using electronic mail transfer was the usual route to receive and organise the data. This proved tobe generally better understood and straightforward to implement. Cross-platform compatibility and a simple modus op-erandi render this tool useful for storage and transportation of images and static data. Other data (animations), had to becompressed separately and transferred across file-systems. The organisation of the MMS-data, was performed in a similarmanner. The related MMS-section had to be identified. This was effortlessly achieved by following the menu choices:first, to the appropriate vessel orientation, followed by the stirrer complexity, and lastly, the particular related variable tothe new information supplied. The next step was to identify secondary reference links. The graph-structure was used ex-tensively to determine links between the various sections of the MMS. At this point, the nature of the data determinesthe links to other frames. Testing and cruise-control links are the final tasks required to complete the system.

Solid-modelling with simulation dataSimulation and solid-modelling visual results often had to be combined for realism. Additionally, in many cases, theexperimental results do not appear in two-dimensional slices; instead, in three-dimensional real-life video-footage. Theaim here is for numerical predictions to be presented in a comparative and realistic manner, similar to that of the experi-ment. One solution is achieved via the use of solid-modelling, with rendering software for the creation of three-dimensional representations, based on simulated data. For this Dough-Kneading case study, a number of solid-modellingtechniques were utilised. As discussed above in figure 22, for free-surface representation, simulated three-dimensionalfree-surface results are combined, with surface visualisation, to generate an animated computer visualisation. This imagewas generated in two stages. Initially, free-surface positions were represented as three-dimensional spherical-particles inspace3. These particles were rendered with the same camera-orientation as that of the solid model. Subsequently, renderedversions were animated as a background to the solid model, for different particle-positions. Finally, resulting rendered-frames were used as a basis to create a full-360° animated-rotation.

Variable combination on-demandOn-demand variable combination is one powerful feature that MMEs introduces, here applied for multimedia content, asopposed to traditional hypermedia applications [9]. Programming such a feature with non-multimedia orientated soft-ware, is generally impractical and time-consuming. Therefore, it was necessary to develop techniques for efficient frame-linking (see Scientific Interactive Multimedia Model/Framework, SIMM/SIMF [26]). The flexibility of such an imple-mentation is more powerful than that achieved with traditional sequential implementations, for complicated presenta-tions. In addition, the lack of control over the multi-media data (including re-timings) requires synchronised streams. In 3 X, Y, Z coordinates were plotted in space as spheres, and animated through viewpoint-rotation using PovRay.

turn, this is difficult to achieve, if streams originate from a variety of sources. The drawback of such an approach lieswith the number of links that must be programmed to support non-sequential frame-access.

The solution to this obstacle, demanding reprogramming of multiple-links, was reached through object-oriented MMframe-creation. Links are separated into two sections: those relating to the global presentation, and those relating only tolocal, current frame-data. The problem of frame-linking appears when a parent-frame is used to create a child MM-frame.Global links must be fully-defined within the parent-frame, before the child-frame global-links are completed, and onlylocal-link programming remains. This enables rapid creation of fully-functional child MM-frames. Global-links are local-ised to the multi-menu construct. Local-links are programmed with the icon-palette at frame-bottom. Although in thisinstance, the menu-location is separate, these may be blended within a single menu.

4. IMPROVING SYSTEM PERFORMANCESystem performance becomes important, with the introduction of multiple, large, simultaneously displayed and synchro-nised data-streams. Presentation delays, slow animation frame-rate and system failures are typical. Investigating howsystem performance can be improved to accommodate such processing strain is critical. Comparison between PPT andMMS technologies when dealing with complex data is required in the corresponding areas of: stream storage/retrieval,stream synchronisation and multiple-stream presentation. Here, large streams are considered. Implementation cost is re-duced, by avoiding hardware-based system improvements [22]. Instead, software-based boosts to overall system perform-ance are presented.

Stream storage/retrievalData-streams can be stored, both locally (hard disk drive) and remotely (via TCP/IP connection) and accessed on-demand,when using both PPT and MME technologies. The choice of processor, speed and type, operating system, size of ran-dom-access memory, bus-size and hard-disk speed can all affect hardware performance. Most of the problems encounteredwithin stream storage/retrieval can be corrected with stream-optimisation techniques, software algorithms for stream-rendering. The result of this procedure generates streams of reduced size, that require less data-transfer per time-unit.Typical, is the example encountered within the MMS [Figure 15]. The two-stirrer, speed-variation section is illustratedusing four animated-streams of different data-types. MB and pressure, animated in two-dimensions, and extension-rate,animated in three-dimensions, through speed-variation. These simulated results are complemented with an experimentaldigitised video-footage of the actual flow for a fixed speed-setting. The original experimental video-stream, in its rawformat, was 115 megabytes in size, without compression. Its total play-length was 30 seconds. The repetitive nature ofthe footage, enabled the animation to be partitioned into smaller sections, from which only an 8 second partition wasselected for the MMS. This first-step reduced the file-size by approximately 1/3 of the original size. Then, a furthervideo-stream compression algorithm was utilised (Indeo 5.1 codec), that reduced the stream to approximately 12 Mbytes.Reduction is achieved to approximately 10 per cent of original size. This enables the MME to load the whole animationwithin system memory and avoid caching using the hard-disk (a very time-consuming operation). Faster response-timesare observed within PPT, when optimised streams are used for similar reasons. The disadvantage with PPT technology isthat, when a stream is encapsulated more than once within a single presentation entity, the original data are stored dou-bly. This increases file-sizes enormously and can cause system failure for relatively simple presentations with multiplestream-instances. This is not the case with MME, that utilise a single-stream. Reference is made to the same, uniquestream, when called from different frames.

Stream synchronisationIn figure 15, stream synchronisation is utilised. This enables the synchronised presentation of different streams in rela-tion with a common changing setting (here speed) [13]. Close observation reveals that motion-blur animations are dis-played and synchronised to those for pressure. When the two animations are played independently, using an externalanimation player, the two streams are of different time-lengths. To set the desired frame-rate for each animation, retimingis utilised. An animation with 60 frames and an animation with 30 frames can be synchronised if the faster animation isplayed at double the frame-rate. Commonly, this is not a practical option, as current realistic frame-rate for personalcomputers (without dedicated hardware), do not exceed 30 frames per second. A remedy is to program the MME to skipevery second frame, so that re-timing is achieved. This approach requires sufficient computation power and relativelysmall-sized streams, else caching will force the MME to drop frames. On the other hand, our favoured option is toachieve the same effect by removing every second frame from the faster animation, an operation reducing the stream-sizeby one-half.

Performance can be improved further at the interface level. Avoiding overlapping animations enables direct-drawing tech-niques to be utilised. Also, with the introduction of multiple streams, using common data-format, faster decompressionis accessed as the same algorithm is used to process all streams. In the main, algorithmic improvements are sufficient toimprove overall system performance. This is a key factor for the distribution of MMS technology. The wide availabilityof the ‘so-called’ personal computer, with basic multi-media capabilities as standard, is sufficient.

5. CONCLUSIONS AND FURTHER DOUGH-MMS ENHANCEMENTSA number of enhancements have been incorporated into the MMS at later stages, to cover presentation requirements for awide variety of target audiences. Multi-mode animation is available in “User-specified variable selection”, “Pre-definedorder across main sections (orange cruise control)” and “Exhaustive, in-order navigation (red cruise control)”. Al-

though these navigation options enable multiple-modes of interaction, (where combination is permitted in special cases),tightly specified order may be required. ‘Green cruise-control’ covers this contingency, where the pre-defined order ofslides reflects presentation order. Hence, backward-compatibility is ensured, to more traditional, linear-presentationmethods (PPT). This property alone categorises the MMS as superior to other linearly-ordered presentation methods. Itsets the standards for interaction, prioritised in terms of data or user-selection.

Beyond the implementation of a pre-defined order, a further iterative step, is the compilation of a voiceover. Digital re-cording and partitioning of the voiceover allows its incorporation into MM-frames. This, in turn, allows for dynamicdemonstration with vocal explanation, relating to each individual frame. Figure 33 shows a characteristic frame that ispart of the green cruise-control path, with VCR slide-bar controller for the digitised voiceover.

Other enhancements include the delivery over different media. Internet delivery has been implemented efficiently throughSHOCKWAVE [25]. The interface is fully-functional and streaming with compression is utilised to deliver multimedia-content. This client-server data-on-request approach, does not require the complete MMS contents to be downloaded be-fore viewing. A few additional improvements may readily be incorporated. Using the client-server Internet content-delivery approach, one may specify data to be accessed from different locations. Data may be stored locally. Once up-dated, the MMS will access and present the latest version. This non-centralised approach is particularly useful when thesubject-matter is on-going and requires frequent updates. It follows the PREMO standard [11], which specifies data asnon-centralised. Note, this does not guarantee all connections will be accessible, and may lead to lack of data for presen-tation.

References[1] A. Baloch and M. F. Webster. Distributed parallel computation for complex rotational flows of non-Newtonian

fluids. Eng. Computations. (submitted), 2001.

[2] A. Baloch, Grant P.W., M. F. Webster. Parallel computation of two-dimensional rotational flows of viscoelas-tic fluids in a cylindrical vessel. Int. J. Num. Meth. Fluids, in press, 2002.

[3] E. Biersack and W. Geyer. Synchronized delivery and playout of distributed stored multimedia streams. Multi-media Systems 7: 70-90, Springer-Verlag, 1999.

[4] M. A. Couch, D. M. Binding, K. S. Sujatha, M. F. Webster. Experimental and numerical simulation of doughkneading in filled geometries. J. Food Eng. in press, 2002.

[5] J. Dakss, S. Agamanolis, E. Chalom, V. Michael Bove, Jr. Hyperlinked Video. Proc. SPIE Multimedia Sys-tems and Applications, v. 3528, 1998.

[6] I. Deliyannis and M. F. Webster. Multi-Level Modelling and Interactive Multimedia presentation of ScientificData over various Media. IADIS 2002, Lisbon Portugal. Also available as CSR 20-2002 at http://www-compsci.swan.ac.uk/reports/, University of Wales Swansea; Computer Science Department, Wales, UK, 2002.

[7] I. Deliyannis and M. F. Webster. Multimedia environments and interactive scientific presentation of industri-ally-based data sets. available as CSR 13-2002 at http://www-compsci.swan.ac.uk/reports/, University of WalesSwansea; Computer Science Department, Wales, UK, 2002.

[8] D. Ding and M.F. Webster. Three-dimensional Numerical Simulation of Dough-Kneading. "Rheology 2000 -XIII Int. Congress on Rheology", eds. D.M. Binding, N.E. Hudson, J. Mewis, J-M. Piau, C.J.S. Petrie, P.Townsend, M.H. Wagner, and K.Walters, BSR, Glasgow, 318-320, 2000.

[9] I. Heath, W. Hall, R. Crowder, G. Wills, J. Ballantyne. Towards a New Authoring Methodology for Large-Scale Hypermedia Applications. Multimedia Systems and Applications, Vol. 12, 2/3, 2000.

[10] J. P. Hughes, T. E. R. Jones, P. W James. Experimental and numerical simulations of in-container mixingprocesses in axial and end over end rotations. Conf. Industrial Food Processing: Experiments and numericalsimulation, Conference Talk. Univ. Plymouth, UK, 2001.

[11] ISO Standard, ISO/IEC 14478 Information technology - Computer graphics and image processing - PresentationEnvironment for Multimedia Objects (PREMO), 1988.

[12] P. W. James, J. P. Hughes, T. E. R. Jones. Numerical and Experimental Simulation of the Flow in RotatingContainers. NAFEMS Int. J. CFD Case Studies, Vol 3 July Pages 31-60, 2001.

[13] W. Liao and Victor O.K.Li. Synchronization of distributed multimedia systems with user interactions. Multi-media Systems, 6: 196-205, Springer-Verlag, 1998.

[14] J. Nielsen. Designing Web Usability: The Practice of Simplicity. New Riders Publishing,

[15] J. Nielsen, (Ed.). Coordinating User Interfaces for Consistency. Morgan Kaufmann Publishers, San Francisco,ISBN 0-12-518400-X (1989-2002).

[16] J. Nielsen and A. Aboulafia. Designing User Interfaces - The Role of Intuition and Imagination. In AdjunctProceedings of INTERCHI '93, pages 209-211, ACM Press, 1993.

[17] K. Rothermel and T. Helbig. An adaptive protocol for synchronizing media streams. Multimedia Systems 5:324-336, Springer-Verlag, 1997.

[18] K. S. Sujatha, D. Ding, M. F. Webster. Modelling three-dimensional mixing flows in cylindrical-shaped ves-sels. ECCOMAS-2001, Swansea, UK. Int. J. Num. Meth. Fluids, 2001.

[19] K. S. Sujatha and M.F. Webster. Modelling Free-surface Flows in Part-Filled Rotating Vessels: Vertical andHorizontal Orientations. J. Fluids Eng., (CSR 2-2002), 2002.

[20] K. S. Sujatha, M.F. Webster, D.M. Binding, M.A. Couch. Modelling and Experimental Studies of RotatingFlows in part-Filled Vessels: Wetting and Peeling. J. Foods Eng., 2002.

[21] H. Thimbleby. User Interface Design. ACM Press, New York, 1990.

[22] P. K. C. Tse and C. H. C. Leung. Improving multimedia systems performance using constant-density recordingdisks. Multimedia Systems 8: 47-56, Springer-Verlag, 2000.

[23] M. Vazirigiannis, Y. Theodoridis, T. Sellis. Spatio-temporal composition and indexing for large multimediaapplications. Multimedia Systems 6: 284-298, Springer-Verlag, 1998.

[24] M. D. Wilson. Enhancing Multimedia Interfaces with Intelligence. Multimedia Systems and Applications. Aca-demic Press, 1995.

Book/Thesis

[25] B. Schmitt. Shockwave Studio: Designing Multimedia for the Web. O’Reilly and Associates, Inc, ISBN:156592231X, 1997.

[26] I. Deliyannis. Interactive Multi-Media Systems for Science and Rheology, Ph.D Thesis, UCS, UK; Universityof Wales, Swansea, 2002.